I clustered four Framework Mainboards to test LLMs

jeffgeerling.com

50 pointsbobajeff6 comments

Discussed heavily yesterday: https://news.ycombinator.com/item?id=44827862

Comments moved thither. Thanks!

Do you have a filter that pops up any links back to HN in the comments to find these mentions of previous similar posts?

I have keyboard shortcuts to quickly drive through HN Search and open/jump around the relevant HN threads and make lists of the interesting ones and then auto-format the list.

(This comes up a lot - see https://news.ycombinator.com/item?id=40564558, https://news.ycombinator.com/item?id=35668525 if curious)

I read the title and thought that sounds like something...saw the domain at the end, yup. Exactly.

> First of all, AMD's support for their AI stuff is... okay. But not great. And that means I spent more time in my testing debugging ROCm driver and library issues than actually testing.

Why the hell can't AMD get their act together? The continued software issues are costing them A LOT of money. It simply must be the lowest hanging fruit toward increasing their market share and revenue.

Reply on news.ycombinator.com